Robust Infrared Small Target Detection via Temporal Low-Rank and Sparse Representation

Infrared small target detection is still one of the key techniques in the infrared search and track systems. We proposed a robust and efficient detection method by exploiting low-rank and sparse representation. We extend traditional low-rank and sparse representation to temporal domain. Initially, we use the proposed method to locate the suspected position of a target in the first frame. Then, we shrink the detection region in local area by considering the fact that the target moves with small distance in the neighboring frames. Finally, we can extract the target trajectory by detecting the image frames iteratively. The proposed approach is tested on several infrared image sequences and compared with the classical target detection methods. The results show that our approach has good detection precision in different image sequences and also achieves better time efficiency than other methods.